Diagram

Three designs for a coding agent's memory

claude-mem captures session events with hooks, has a model write observations, stores them in SQLite and Chroma, and adds context at the next session start. MemPalace saves transcripts word for word through hooks into ChromaDB and recalls by search. XERJ waits for the agent to call store or recall over MCP and keeps memories in a local index. claude-mem Hooks capture Model summarizes SQLite + Chroma Context at next start MemPalace Hooks save Word for word ChromaDB MCP search XERJ Agent calls store Text as sent Local index MCP recall BM25 / hybrid
claude-mem and MemPalace save on their own through hooks; XERJ saves only when the agent calls a tool. claude-mem has a model write summaries; MemPalace and XERJ keep the original text. Read from each project's README on 2026-10-30.